Introduction to Archaeology: Class 9
Finding sites and reading culture from maps
Copyright Bruce Owen 2002
if it would be too slow or expensive to cover the whole landscape, some kinds of questions can be answered by studying just a sample of the area
- Finding sites: survey methods
- archaeologists usually find or confirm sites by recognizing artifacts on the surface
- sometimes walls, foundations, or mounds of accumulated debris are visible
- more often, bits of broken pottery or stone tools are scattered on the surface
- more details on this later...
- many sites are obvious or were never lost
- hard to miss the pyramids, Roman coliseum, etc.
- asking the local people
- they often know where they have found artifacts while plowing fields, digging wells, etc.
- they may have oral traditions of where events occurred in the distant past
- old maps and documents
- air and satellite photos
- may show subtle patterns in relief
- soil color
- plant growth, etc.
- accidental discovery
- such as when a bulldozer working on a construction project plows up some artifacts
- very common; many important sites have been found this way
- The most thorough method is systematic site survey
- walking in an orderly pattern back and forth across the landscape and recording all artifacts that are laying on the ground
- If there is ground cover or sites are likely to be buried, it may be necessary to dig small trenches in likely areas, bore holes with augers, etc.
- Systematic survey is slow and expensive
- but unlike the other methods, it can give you a more complete, relatively unbiased picture of the distribution of sites in the whole region
- not just the places that happen to have standing architecture, be remembered, have been discovered accidentally, etc.
- Thomas emphasizes looking for particular sites - a cave that will answer his questions, or a specific historical mission
- But often we just want to know what is there in general, rather than find a specific site
- Or we want to look at the settlement pattern: the pattern of distribution of sites of different types across the landscape, and their relationships to natural resources and each other
- The settlement pattern of a given region and time can tell us a lot about the culture that produced it - often things that would be difficult or impossible to learn from excavating a single site
- simply finding surface evidence of sites, recording what is there and where, can provide a huge amount of information for much less effort and cost than excavation
- 100% coverage (or "full coverage") survey methods
- line of archaeologists walking across the countryside, looking at the ground
- spacing within the line and between passes determines the largest possible site that could be missed
- collect standardized information about each place where any evidence is found
- field forms
- location, site number or name, how to find it if necessary
- description of sherds, lithics, walls, etc. found
- size of surface scatter of artifacts; map of the area covered or any architecture
- description of the place: location, topography, ecology, access to water, travel routes, view, etc.
- surface collections
- "grab sample": purposefully collected items that seem interesting or diagnostic
- "systematic sample": collection made by systematic rules. For example:
- Draw a 2 meter diameter circle on the surface near the center of the scatter, using two large nails tied together by a 1 meter string
- Pick up every sherd within the circle that is larger than 3/4" square
- and/or every stone tool or flake of stone tool material larger than 1/2" square; or whatever
- These are taken back to the lab, categorized, cataloged, and used to date and otherwise characterize the site
- some environments where you can't see the ground due to plant cover or where artifacts are typically below the surface may require:
- periodic shovel tests, post-hole samples, or auger samples
- periodic clearing using a machete, etc.
- this is VERY slow and expensive to do over a whole landscape
sample areas are chosen to test areas of interest
like: one area on the shore of each lake; one on each hilltop; etc.
this may introduce biases into the data
- like a series of 100 x 100 meter areas scattered across the landscape
- this could help with questions like:
- what is the general density of different kinds of sites?
- what are the common kinds of sites, and what are the rare ones?
- what kinds of areas are sites typically found in?
- near lakes, on hilltops, or in meadows?
- this could tell you a lot about the subsistence practices of the people who lived at the sites
- or their need for defense, etc.
- are certain kinds of sites found only in certain parts of the region?
- did the red-pottery people live to the south and the white-pottery people to the north, or were both types of pottery used throughout the whole region?
- this could tell you about ethnic groups, political relations, economic activities, etc.
sample areas are selected in some regular pattern
like 100 x 100 meter square areas arranged in a 1000 meter grid
- sites in areas that the investigator does not think are worth checking will never be found - his assumption will never be disproved
- the resulting data does not accurately represent the region as a whole
- you couldn't say things like "there are 2 sites per square kilometer", because you have only looked in certain places; who knows what the site density is in other places?
this is less biased, but could still have problems.
- this would be a 1% sample of the area
- probably a bit small for most purposes, but a start
sample areas are selected using a random number table, random number generator on a computer or calculator, etc.
should avoid introducing any systematic biases into the data
all parts of the region are equally likely to be sampled
spatial patterning of site locations will not affect how likely sites are to be found
the final data is a fair representation of the whole region
stratified random sampling
but why waste lots of time looking in the unlikely areas where you never find anything?
or why risk missing certain parts of the region completely?
- say sites tended to be about a kilometer apart.
- with luck, you might hit every site, and think the site density was very high
- or the sample areas might fall in between them, and you would think the site density was very low
solution: stratified random sampling
- say only 0.1% of the region is riverbanks, but you think people probably liked to live along riverbanks
- a random 5% sample might well not include a single survey area that falls on a riverbank
then do a random sample within each stratum
- divide up the region into a few kinds of terrain, like "exposed bedrock", "hillsides", and "riverbanks"
- these are the "sampling strata"
this gives you a truly representative, unbiased sample of each kind of area
you can extrapolate this to the entire area by multiplying by the area of each stratum
- this assures that you cover some of each
- you can even do a higher percentage sample in the more promising areas
- say, just 1% coverage of the exposed bedrock areas
- but 10% coverage of the riverbank areas
but random sampling areas are often hard to get to; they require lots of wasted travel time
- say you find 0.1 site per square kilometer in the exposed bedrock sampling stratum
- 1 site per square kilometer in the hillside stratum
- 10 sites per square kilometer in the riverbank stratum
- to estimate the total number of sites in the region, you use a map to measure how many km2 each stratum occupies
- then multiply the site density in each by the area of each
- this gives you an unbiased estimate for the whole region without having to cover too much exposed bedrock...
these same sampling strategies will be useful in laying out surface collection units on large sites, selecting spots to excavate, etc.
problems with sampling:
- transects are sampling units in the form of long strips
- the strips typically cross a variety of terrain types
- so they tend to give a fairly representative sample of the landscape
- but they are easier to cover without a lot of wasted travel time
- transects are not really random and unbiased, but they are a fair approximation
Settlement pattern analysis
generally works best with 100% coverage, so patterns can be seen
Gross patterning: what kinds of sites are where?
- by definition, you only get data about part of the area
- most of the area is never covered
- so, rare or unique types of sites can easily be missed
- if you did a random sample survey of lower Egypt, odds are high that you would miss the great pyramids... and that would certainly mess up your conclusions.
- also, you don't get the whole pattern, just little pieces of it
- so a sampling scheme would tend to hide spatial patterns like:
- sites were located in lines along rows or canals
- sites arranged with a few large sites, each surrounded by a halo of small sites
- patterns like these could be very useful for reconstructing social organization - but sampling techniques rarely reveal them
Catchment analysis: what resources were sites located near?
- are there sub-regions associated with different kinds of sites (farming villages vs. hunting camps; different ethnic groups, etc.)?
- did the preferred locations of sites shift over time?
- due to changing environment? economic specializations like fishing vs. farming? building of canals? etc.
- did overall population rise, fall, or remain constant?
- are sites associated with certain artificial features, like roads or canals?
- this can help to date otherwise enigmatic features
- and to suggest what they were for
- this level of settlement pattern analysis can often work with sampled data
Site size hierarchies: do site sizes reveal something about political organization?
- simply draw a circle of "x" km radius around each site on a map that shows ecological zones or other distributions of resources
- measure the area of each kind of resource zone within the given distance from the site
- this gives a rough idea of what resources were probably important to the people at the site
- slightly better way:
- draw the boundaries according to walking time, rather than simple distance
- some terrain may be harder to cross, or there may be natural barriers like rivers or cliffs
- what if the circles intersect (as they probably will, unless the region is seriously underpopulated)?
- use Thiessen polygons: a way of dividing up the landscape into areas that probably "belonged to" each site
- simplest way:
- draw lines connected each site to every other nearby site
- draw a perpendicular line bisecting each of the connecting lines
- these form polygons such that the area within each is closer to the site in the polygon's center than to any other site
- slightly more sophisticated way:
- estimate the size or population of each site
- rather than bisecting the connecting lines, divide the connecting line proportionally to the size or area of the sites (farther away from bigger sites, closer to smaller sites)
- there are computer programs for doing this sort of work
Nearest neighbor analysis: a statistical analysis that gives an estimate of the randomness of site locations
- sites all the same size would suggest an unspecialized society in which all settlements had similar functions
- a few large sites and many small ones would suggest that the large sites might have had some additional functions that the small ones didn't
- like a temple, royal palace, marketplace, craft workshops, etc.
- even without any other information, such a pattern of site sizes suggests a more complex social, political, and/or economic organization
- usually presented as a "rank-size" graph
- with site size on the vertical axis (usually logarithmic)
- and site size "rank" on the horizontal axis (that is, first on the left is the largest site, next is the second largest, etc.
- the shape of the curve that this forms gives an idea of the distribution of site sizes
- a steep drop from a single very large site is a "primate site size hierarchy": one big capital-like site, probably a highly centralized, strongly administered society
Analysis based on Central Place Theory: looks for patterns of administrative or economic centers spaced widely apart, each surrounded by smaller sites dependent upon them
- why on earth would that be useful?
- the assumption is that natural resources are more or less randomly distributed, or are evenly distributed
- so if sites are randomly distributed, their locations are not affected much by the locations of other sites
- so social interactions between sites, competition, administration, etc. were not very important in determining their locations
- but if sites are very non-randomly distributed, that means that the average distances between them tend to be somewhat regular
- that is, the location of one site is partially determined by the location of others
- what could cause that?
- competition: the sites are evenly spread out to maximize the area each has access to and minimize closeness to others
- administrative needs: sites are neither too close (which would waste administrative resources due to duplication) nor too far apart (which would stretch lines of communication and make administration more difficult)
- trade: sites are neither too close (not enough market for their goods) nor too far apart (excessive transport costs for goods)
and there are many other more specialized approaches
Point: from just in the spatial patterning of sites and the small amounts of material on the surface, one can figure out a lot about past societies
- ideally in roughly hexagonal patterns - but this would only be so on the proverbial "featureless plain", never in reality
- may have more than two levels of site types
- if there is such a pattern, it indicates even more complex, centralized administration or exchange